107 replies to this topic

[piaras]

I was thinking of mono not 3 color cameras. If one spends 2k for a mono with an APS-C chip even at Black Friday prices where is the savings? 

Edited by piaras, 24 November 2023 - 09:03 PM.

[Michael Covington]

It may not come out cheaper, but the gap is much narrower than it used to be.  The very economical astrocameras have sensors smaller than APS-C, such as the 533 and 585 sensors.  The sensors are smaller but they certainly take good pictures.

Edited by Michael Covington, 24 November 2023 - 09:34 PM.

[unimatrix0]

From my elementary study of pages like "photos from photons" (or something similarly called), I sort of got the idea that many camera manufacturers doing all sorts of pre-processing tricks - which probably takes a fraction of a second- to manipulate the final image outcome to be the least noise free; more vibrant and earn a higher score on the benchmark tools, so they can say that "our camera has gotten a score of 9.4, while the XY camera got 9.1".

Nevermind the fact, that they seem to concentrate on people taking .jpegs so somehow they manipulate the shady in-camera-processed .raw to have the best converted .jpeg image to come out. 

It's most likely the least of their concerns if someone try to use it for astrophotography, because that makes about 0.01% of the potential buyers.  

 

So just my opinion, that extreme low light photography, such as DSO photography that requires stacking images is not their concern. The focus is on having a camera produce the best looking (doesn't even have to be realistic) .jpeg to come out from the camera on the "auto" settings, with a zoom lens. 

[Michael Covington]

Does the Canon EOS Ra have the same rings and gradients as the EOS R?  If so, that's really bad news, because the Ra is sold specifically for astrophotography.   There's also a Nikon astrophotography DSLR (D810a, I believe it is), which I've never used, and two previous Canon ones, 20Da and 60Da, both of which I've used.   

The 20Da and 60Da are free of rings and other shenanigans as far as I can tell.  Reportedly, the Nikon D810a is. But those are not low-priced cameras.  Nor is the Ra.

[sharkmelley]

The 20Da and 60Da are free of rings and other shenanigans as far as I can tell.  

You say the 20Da and 60Da are free of rings but have you have carefully examined the histograms at various ISOs?

 

Yes, the EOS Ra does have these rings:

https://www.cloudyni...ing/?p=10731019

However, on that individual camera they disappeared at ISO 800 and above. Unfortunately I don't have raw files above ISO 100 for any other EOS Ra. 

 

As I said earlier, this problem (histogram gaps/spikes and their associated rings) is caused by digital scaling using a multiplier very close to unity.  Evidence of Canon applying digital scaling goes all the way back to the Canon 6D but back then it was much more benign i.e. the multiplier was not so close to unity.  Also, not every 6D shows evidence of these histogram gaps so it could be some factory calibration that causes it, as suggested by Iliah Borg who authored LibRaw. 

[Michael Covington]

I would be glad to test my 60Da for histogram gaps -- what is the best way to do that?  I have never had artifacts in images other than a slight striped pattern in both directions ("the tartan of the clan of Canon") which is not a problem when I calibrate in the normal way.

I no longer have the 20Da, which was on loan from Canon way back when.

Edited by Michael Covington, 25 November 2023 - 10:01 AM.

[sharkmelley]

I would be glad to test my 60Da for histogram gaps -- what is the best way to do that?

A couple of flats - one at ISO 100 and another at ISO 1600 should do the trick

[Michael Covington]

Here's a histogram of a flat taken with my 60Da at ISO 640, the setting I normally use.  (640 is the lowest of three ISO settings that seem to have the same read noise; see photonstophotos.net.)  I don't see anything fishy here; do you?

Maybe this is the perfect camera and I should stick with it!  As you know, the bugaboo with astrocameras right now is that people are getting vibration from the fans... much unit-to-unit variation... at least the DSLR doesn't have a fan.

[Michael Covington]

Same with linearly scaled vertical axis (to show that the small irregularities in the first picture are due to sampling).  Again, Canon 60Da, ISO 640.  I'm sticking with this...

[gregbradley]

Astro dedicated cameras have dropped, but what about the filters, filter wheels, and 4x7 times the exposures required to get a photo? I live in an area that is not conducive to doing this, so I will stick with my DSLR until circumstances change.

Pierre

There is a complete range of one shot colour dedicated astro cams. They are quite popular. No need for filter wheel and filters (although there are purpose made filters to enhance some of the emissions we image).

 

Some of the lesser models are quite a bit cheaper than mirrorless cameras.

 

Greg.

[Michael Covington]

Right, we should compare apples to apples.   A DSLR or mirrorless camera has a low-noise sensor (probably APS-C) and no cooling.

Here is one with a sensor the size of a quarter of an APS-C sensor (that is, half the size in each dimension), and no cooling, for just $360:
https://www.astronom...ing-camera.html

Most of us would want a bigger sensor.  Here is a 1-inch-square sensor, very low noise, with cooling, about $750:
https://www.astronom...phy-camera.html

If you insist on a full APS-C field, here is the camera a lot of us really want -- APS-C sensor and also a guiding camera built in.  With cooling of course.  $1800.
https://www.astronom...r.html?___SID=U

By comparison, a *used* Nikon D5300 with H-alpha conversion is $634 at Lifepixel.  Note that that's a used camera, and it's one of the few Nikons that do not have the mysterious ring problem when not attached to a Nikon lens, but you still need to stay below a certain level to avoid compression artifacts.  And it has no cooling.  The Canon EOS Ra is no longer on the market but was about $2500 new as I recall.  

[sharkmelley]

Here's a histogram of a flat taken with my 60Da at ISO 640, the setting I normally use.

Those 60Da histograms look perfect!  No evidence of digital scaling.

 

Compare to this histogram of a Canon 6D flat where the histogram gaps are quite obvious:

 

 

I should point out (again) that not all Canon 6D cameras show the same behaviour.

[ErwinL]

Compare to this histogram of a Canon 6D flat where the histogram gaps are quite obvious:

Strange. Did you ever see those gaps at ISO 800 or 1600 ?

[Michael Covington]

Here is the graph on PhotonsToPhotos that led me to choose 640...

 https://photonstopho...non EOS 60Da_14

 

Edited by Michael Covington, 25 November 2023 - 06:30 PM.

[Michael Covington]

Note also that my RawDigger graphs are binned 2, not 1.   More testing awaits, with a less generously exposed flat.

[sharkmelley]

Strange. Did you ever see those gaps at ISO 800 or 1600 ?

Yes the gaps exist at ISO 1600 as well as ISO 100.  Here's a histogram of a 6D light frame at ISO 1600:

 

[Michael Covington]

OK, with the 60Da it gets interesting.  Using a less generously exposed flat, so I can bin the columns by 1 (i.e., not bin them), there are still no GAPS but there are SPIKES (stegosaurus fins?) especially at 640.  That suggests 640 is a digital compression of 800 and I should be using 800 from here on.  Your thoughts?

[sharkmelley]

Yes, you are seeing digital scaling at ISO 640.  It's been known for a very long time that on some Canon cameras the intermediate ISOs between 100, 200, 400, 800, 1600, 3200 etc are achieved by digital scaling of the "native" ISO values.  This scaling doesn't cause the ring-like artefacts under discussion here because the scaling factors are not sufficiently close to unity, so the histogram gaps/spikes are not widely spaced.  More technically, the read-noise dithers the data and reduces the amplitude of the potential rings to a point where they are impossible to detect, even with stacking.

[ErwinL]

Yes the gaps exist at ISO 1600 as well as ISO 100.  Here's a histogram of a 6D light frame at ISO 1600:

Might the appearance of the gaps depend on the firmware version?
 

[sharkmelley]

Might the appearance of the gaps depend on the firmware version?
 

It's certainly possible that the Canon 6D histogram gaps depend on firmware version rather than factory calibration.  However, we would need many more examples to come to a definite conclusion.

[ErwinL]

The reason for my question was that I vaguely remember that the Magic Lantern full-resolution silent pictures of a 60D had gaps, while the CR2's did not.

I'll dig through my data to see if I can still find it.

[Michael Covington]

Yes, you are seeing digital scaling at ISO 640.  It's been known for a very long time that on some Canon cameras the intermediate ISOs between 100, 200, 400, 800, 1600, 3200 etc are achieved by digital scaling of the "native" ISO values.  This scaling doesn't cause the ring-like artefacts under discussion here because the scaling factors are not sufficiently close to unity, so the histogram gaps/spikes are not widely spaced.  More technically, the read-noise dithers the data and reduces the amplitude of the potential rings to a point where they are impossible to detect, even with stacking.

Thanks!  I had assumed that since 640, 800, and 1000 form a group with about the same read noise in electrons (and then 1250, 1600, and 2000 form the next group, and so on), the "native" ISO was the lowest one in the group.  Now I see it's the middle one in the group.  So I'll re-standardize at 800 and 1600 as my usual ISO settings.  I spent some time last night taking dark frames -- our outdoor temperature was very stable and was at a level that matches a good many of my observing sessions (48 F = 9 C, which gives me a stable sensor temperature of 22 C).  The one thing cloudy weather is good for is that it gives you temperatures that don't change for a couple of hours.

[ErwinL]

Hi Mark,

 

here are the 60D frames I looked for. The first (log) histogram shows a normal dark image. There are strong dips, but this is mainly due to the fact that it is a dark and the exposure time is only 16 seconds, which is the limit for full-resolution silent (FRS) pictures.

 

The second histogram shows an FRS frame taken under the same conditions. It shows additional gaps down to zero with one third of the CR2 dip frequency.

 

What can't be seen here are about 20,000 dead pixels and a similar number of warm and hot pixels which do not appear in the CR2 any more.

 

From these observations, my guess is that FRS frames come directly from the sensor with gaps introduced by A/D hardware.

On it's way to the CR2 frame (through the digic processor) the histogram is smoothed, perhaps similar to what you did with your EOS R frames.

 

What do you think?

[ErwinL]

I had assumed that since 640, 800, and 1000 form a group with about the same read noise ...

I think this assumption is correct. However, the Canon ISOs are a bit below the ones determined by P2P and somehow there must have occured a shift in the P2P diagram (around ISO 2000 one data point is missing).

Edited by ErwinL, 26 November 2023 - 10:57 AM.

[sharkmelley]

From these observations, my guess is that FRS frames come directly from the sensor with gaps introduced by A/D hardware.

On it's way to the CR2 frame (through the digic processor) the histogram is smoothed, perhaps similar to what you did with your EOS R frames.

 

What do you think?

Your results are not what I would have expected.  I would have expected that Magic Lantern would switch off the digital scaling that causes the histogram gaps.  

 

What you suggest is certainly possible but it would seem very weird that Canon's sensor is applying digital scaling which is then reversed out by the Digic processor.   It's all a bit mysterious.

 

In any case, the gap spacing in that histogram is approximately 50 and that is definitely in the danger zone for creating concentric ring artefacts in stacked images.

 

Mark